Fig. 2: scAGDE is the most of the compared clustering methods on varying simulated scATAC-seq scenarios. | Nature Communications

Fig. 2: scAGDE is the most of the compared clustering methods on varying simulated scATAC-seq scenarios.

From: Topological identification and interpretation for single-cell epigenetic regulation elucidation in multi-tasks using scAGDE

Fig. 2

a Comparison of ARI values obtained for each method across five simulated datasets of varying sequencing depth. A value approach 1 indicates optimal clustering. The sequencing depth is represented by fragment numbers per cell, which covers a low depth of 250 to 500, moderate depth of 1500 to 2500 and high depth of 5000. b Table displaying NMI average results between scAGDE and compared methods under five noise levels. c UMAP visualization annotated by the true cell-type labels (first column, True-label) and obtained clustering labels from each method (subsequent columns). The first to last rows, respectively show clustering on simulated datasets with low (10%) to high (40%) noise levels. d Table displaying NMI values for each method across seven simulation datasets with varying dropout rates (top, 10–70%). The violin plots (bottom) aligned with the top horizontal axis displaying the increasing sparsity distribution of corresponding datasets (n = 1200 cells in each group; center black dot, median; box limits, upper and lower quartiles; whiskers, the 95% confidence intervals; violin plot edges represent minima and maxima). Source data are provided as a Source Data file.

Back to article page